[Megathread] DotA 2

[Charge me Doctor]

I found a checkbox of "only solo que" in the prefered language thingy. Probably some of you will find it useful

[Deleted]

80k

if i had known. i would have given keys to bots to hand out. i got only 1. atleast got it yeaa.

also why is change notes so obscure..
abaddon and this was a pleasant suprise

[Longview]

Cue Longview linking the post about how your 'replay bracket' has nothing to do with your MMR for the 15th time.

I'm just going to lift the entire post I did last time, you feel me?

Here's a few points I made a while back, they're relevant to the MMR discussion.

There has been statements by high-ups over at dev.dota2.com that levels are purely cosmetic, and they've also said that win/loss-ratio doesn't determine what your rating is (neither does number of wins). It is not rare to see a difference in the 40-50's when I queue.

The only thing we do know about the match making system is that it's a modified version of TrueSkill, which is more accurate than ELO in a multiplayer scenario. ELO works better for single player scenarios such as chess. Read more about TrueSkill here.

We also know that match making rating is not based on wins or losses, and we know that there are a lot of variables in action. What variables this might or might not be is pure speculation unless you are an employee of Valve. If you ever visit Valve HQ, they may tell you what percentile you're in - they told TobiWan that SingSing was the highest ranked player when he visited. That was a long time ago though.

[...] the three tiers of rating are not related to match making rating. Rather, they're there to help distinguish "good" games with higher than average G/XPM and whatever else Valve chose - with the intent to help players find good replays. (Source)

Here's some Valve quotes on matchmaking if you're still trying to wrap your head around it:
http://dev.dota2.com/showthread.php?p=282075#post282075
http://dev.dota2.com/showthread.php?p=282033#post282033
http://dev.dota2.com/showthread.php?...854#post156854
http://dev.dota2.com/showthread.php?p=156150#post156150

TLDR: MMR isn't based on win/loss (hi hermanni), has nothing to do with your cosmetic levels, or the number of games you have played. There's a ton of information being pulled from every player in each game to determine your ranking. You're then queued against players at your own rating (the longer the queue, the broader the potential MMR difference between players in one game becomes). If you're getting feeding russians every other games, this means two things:

• You've a case of selective memory
• These players are at or around your own skill level

One thing i miss from HON was it's ranking & matchmaking/balancing system. People always raged about it, but it worked like a charm. In dota2, the moment i enter a very high game rating, i instantly get matched with feeding russians (MID OR FEED PIZDA SUKA) and plop back down. Sigh.

I disagree. I live with two dudes who still play HoN (and Dota 2, but mainly HoN). Shit like "lmao dont let brown or oj go mid fucking shitters" is commonplace. Repulsive.

That's why I'm situated between High and Very High right bud?

In HoN, the equivalent of being in "high" matchmaking is around 1600 rating, so probably, yeah. That's true if a majority of your games are there.

The system is broken.

Do you have anything to back this up with? Anything at all except guesses and critique? No? We've had this discussion well over ten times, and it's always the same thing being posted. Contrary to your experience, I do not find that the system is broken. Very few times over the course of 1500-something games have I felt that the skill difference of the teams are that bad. It's a fruitless and redundant discussion, nothing will come of it and nobody will agree - unless you have something to stand on.

[Deleted]

Dotabuff had for a short while a rating system implemented until Valve decided to make data private. According to Gabe Newell, the system Dotabuff used was very similar to the one that was implemented in Dota 2. The reason Valve haven't made ratings public, according to Gabe, is because they haven't implemented a system which they believe is good yet.

[Longview]

Dotabuff had for a short while a rating system implemented until Valve decided to make data private. According to Gabe Newell, the system Dotabuff used was very similar to the one that was implemented in Dota 2. The reason Valve haven't made ratings public, according to Gabe, is because they haven't implemented a system which they believe is good yet.

source plz

the only thing I've seen is where he said that their system was shitty

[Deleted]

source plz

the only thing I've seen is where he said that their system was shitty

http://www.youtube.com/watch?v=jhgOqyZHBIU

19:57

[Deleted]

http://dotametrics.wordpress.com/201...s-not-a-carry/

Just read the url.

I remember people talking about how Abaddon should be a carry instead of support earlier in the thread, so thought I'd post this.

[Hermanni]

http://dotametrics.wordpress.com/201...s-not-a-carry/

Just read the url.

I remember people talking about how Abaddon should be a carry instead of support earlier in the thread, so thought I'd post this.

The writer seems to be stuck on the idea that the oh em gee number 1 safe lane trilane farmer is the only way to play carry. And that you need a steroid to be able to carry. And every hero only has one 'real' role. And that the current pro 'metagame' is the absolute only way to play Dota. All in all, a lot of words over a mindset that is already too narrow to discuss the issue.

[Deleted]

http://dotametrics.wordpress.com/201...s-not-a-carry/

Just read the url.

I remember people talking about how Abaddon should be a carry instead of support earlier in the thread, so thought I'd post this.

Abaddon works fine as a carry. He also works as support. Why must a hero be stuck in one role? Play the hero according to your hero lineup and enemy heroes.

[Longview]

http://www.youtube.com/watch?v=jhgOqyZHBIU

19:57

cool thank you

I still haven't lost with support Abaddon, 6-0

Zefie support too good I think

wasn't it "never play abaddon support" or something

[Finear]

MATCHMAKING
- People who decline Matchmaking are temporarily restricted from searching for 45 seconds.
- Parties are more likely to be matched against other similarly sized parties.
- When matching parties, the matchmaker now also considers the minimum experience level on both teams rather than just the average.
- Reduced the average skill variance in games where there are 10 individual players.
- Fixed a bug in the matchmaker that was causing it to insufficiently segregate players based on win count.
- Fixed a bug that would cause excessive queue times, especially for large parties or high skill players.

nice to see some changes

[Longview]

new earthshaker model looks dope as hell

[Hermanni]

our mid

Stopped reading at that. If you want to maintain any credibility in your argument against the MM system try to at least come up with something besides "I always play perfect but mid feeds so he must be bad." I've read people write that exact same thing too many times.

In my experience, if your game isn't in the first 20 pages you shouldn't be surprised if someone doesn't fully understand what their hero does or how some abilities interact with each other.

[Longview]

Link? I'm at work.

http://imgur.com/a/gGyjy

Also - Longview if I were to start gathering data would you start to believe me?

I don't know what type of data you could gather that would indicate how good a player actually is. They shut that out of the API.

As an example, I played Weaver today. Here's the first one and here's the second one. I played one more. You can't draw any reasonable conclusions about my skill-level from the stats. It makes absolutely no difference whether I die seven times in a damn row, or if I go 16-1. I'm just as mediocre.

I won't argue the fact that dying seven times in the same lane in a row should be awarded with some sort of moron medal, but that's what I'm saying. You can't get any type of accurate data from K/D/A, number of games played, win-rate or whatever. None.

It sucks to solo-queue, I don't think I've been in a MM-game alone in well over a year by now. Dota is a team game, and it's best played with a stack. I quit HoN mostly due to visible ratings and getting flamed every single game I played because my friend played my account up from 1450 to 1600. There wasn't anything I could do. If I joined a "noobs only" lobby, they'd kick me because my KDA was too high. If I joined a MM-game, I'd get ravaged and flamed. I think people having bad games and getting flamed is awful, but I think that getting flamed due to your visible rating/KDA/winrate/whatever ("don't let brown go mid fucking shitter you better support") is ten times worse. I know that wasn't your point, but I don't see many other alternatives to "fixing MMR".

Personally I haven't noticed the matchmaking system being fucked. This might very well correlate with the fact that I never solo queue.

If you do end up gathering some sort of non-useless data, I'd be interested to see it. If I would believe you or not is a different case entirely.

[Finear]

Link? I'm at work.

http://i.imgur.com/M0sg1iU.jpg

eh too slow


kinda related to abaddon necro tree strat

[Hermanni]

You've read that far too many times because it's so common. Game starts with a guy demanding mid even though his hero isn't suited to playing against enemy QoP mid(another example) and he proceeds to feed, not gank and just generally let QoP get horribly farm.

I've heard it too many times because people like misjudging their skill level and then using anecdotal evidence as an argument for their 'true' skill. I'm 99% certain you're right where you should be in terms of MM and if you aren't you will be within 10 games.

This may come as a shocker but usually the mid in either team loses. Sometimes they just get less farm and sometimes they get completely owned, but that doesn't prove that anyone in the game is higher or lower in MMR. If your queue time goes above 5 minutes or so then you can expect less balanced games, and if you don't solo-queue then you'll have to remember that your party might consist of 2 or more friends of very different skill levels queuing together.

[Longview]

kinda related to abaddon necro tree strat
http://www.youtube.com/watch?v=tMgv69stb7w

hahaha 10 minute rax with 3 - 1 killscore

nice

[Finear]

ive been in beta since november 2011 and i also cant say that mm is broken
early into beta i was pretty rusty after 1.5y of lol my friends were able to play more and difference between solo queue and stack was huge later i managed to catch up and my solo games were much better and as i keep improving (at least i hope so ;p ) my teammates still are at similar lvl of skill unless i plat at like 5am 1 server AR (what hermanni said about queue times)

hahaha 10 minute rax with 3 - 1 killscore

nice

and 5 blademails

poor tinker ;c



damn thats super cool, now you can time rosh from chat wheel

more here
http://imgur.com/a/wXaVS

[Grable]

Matchmaking is seriously buttfucking my team. There is something wrong about it since release. It never worked like this.

To give a more detailed example, our team was this (win count per player): ~150, 94, ~63, ~30, <10 while their team was 200+, 300+, 300+, 400+, 600+

I mean what the hell. And we weren't on a winning streak. And they matched us 3 or 4 games like this, with us having 2 newbie players and 2 players below 100 wins, while they had 200-600 wins. I wouldn't mind if it matched us against people with 100-200 wins (even then it would be unfair because we had 2 *almost complete* rookies), but this was just beyond silly. There is no way we can teach our rookies if we get guys like this that even the most experienced players on our team have problems dealing with.

[Deleted]

Zefie support too good I think

Pressing W on people who eat stuns is hard.

http://imgur.com/a/gGyjy

But why he looked way better before.

It's just grabbing the first other group of 5 it finds and throwing you in the deep end. Seriously broken system.

Did you try the limited hero pool matches in the tutorial? Those are truly first-5-it-can-find. Try queueing for one and if it's exactly the same as every game you get when queueing, well...